Multinomial Regression for Wide Datasets sourced from Cancer Genetics

• src folder contains all source code
• src/data folder contains all data
• results folder contains results in an excel file and a png file

helperFunctions.R handles the following

• Read the data from a database file
  • Format the data for the regression
• Build a multinomial regression model with the lasso
  • Record the regularization constant (lambda) for the model yielding the lowest classification error
  • Record the number of coefficients (# of genes) used by this best model
  • Record the misclassification error from the cross validation
• Use the generated model to predict labels for the same training data
  • Record the training error

testReadData.R serves as the main executable here and performs the following:

• Sets up the workspace
• Reads all the database filenames from a file containing all the names
• Creates a Dataframe to hold all the values that will be recorded for each regression
• Runs the program defined in “helperFunctions.R” on each of datasets
  • Stores the results in the created Dataframe